双层聚类模型在日志数据分析中的应用

提出了一种基于自组织特征映射( SOM)神经网络和模糊c-均值( FCM)的双层聚类方法，对Web日志中的日志数据集进行聚类。第一层是无监督SOM神经网络聚类方法，它所产生的类的个数大大减少了原始数据集的个数，降低了FCM对类初始中心点的依赖；然后利用FCM聚类算法的优势对第一层中产生的类的中心点进行聚类，从而大大减少了聚类的时间复杂度；最后通过平行坐标技术可视化展示聚类前后的日志数据集，方便对日志数据进行分析。

The Application of Double Layer Clustering Model on Log Data Analysis

A double clustering model to make web log data sets clustering was proposed based on the self-organizing map ( SOM) neural networks and the fuzzy c-means ( FCM) method. The first tier uses unsu-pervised clustering method—SOM neural network, so the number of classes it generates significantly re-duces compared with the original data set, it also reduces the FCM method' s rely on class initial centers. Using the FCM clustering algorithm to cluster the center points of classes generated by the first layer, the time complexity of clustering is greatly reduced. Meanwhile, the parallel coordinates visualization tech-nology to demonstrate the log dataset was used, it is suitable to analyze the log data.